The solution is obviously to create $n$ groups for each newly added ingredient in the previous version, randomly assign customers to a group and ship them a ver

The solution is obviously to create $n$ groups for each newly added ingredient in the previous version, randomly assign customers to a group and ship them a version with one ingredient added, then use R's prop.test to analyze the failure rate of the formulations. /s